Method for evaluating open risk of hard road shoulder of interlaced area of expressway based on ITA and KDE

Abstract

The invention discloses a highway interlacing area hard shoulder open risk assessment method based on ITA and KDE. The highway interlacing area hard shoulder open risk assessment method comprises the following steps: step 1, collecting and recording vehicle operation data in different scenes; 2, track extraction and coordinate preprocessing; step 3, identifying potential conflict risks based on ITA; step 4, carrying out space aggregation on conflict points by utilizing a KDE method; 5, comparing the KDE results in the non-open scene and the open scene of the hard shoulder; and 6, judging the critical and saturated intervals of the hard shoulder by combining the ramp / main line flow ratio. According to actual measurement data of the same road section in a peak period, vehicle tracks in a hard shoulder non-open scene and a hard shoulder open scene are respectively collected. Through visualization and superposition comparison of respective ITA distribution and KDE results of two groups of data, on one hand, a main line inner side lane conflict mitigation effect brought by hard shoulder opening can be quantified; on the other hand, risk outward movement caused by the fact that a vehicle crosses an outer lane and a hard road shoulder can be visually observed.

Description

Technical Field

[0001] The present invention belongs to the technical fields of road traffic engineering and intelligent transportation systems, and particularly relates to a method for evaluating the opening risk of the hard shoulder in the weaving section of an expressway based on ITA and KDE, which is used to quantitatively evaluate and spatially identify the potential traffic conflicts and risks generated before and after the opening of the hard shoulder in the weaving section of the expressway. Background Technique

[0002] With the continuous development of China's economy and the continuous increase in the vehicle ownership, the expressway often appears seriously congested during peak hours in the morning and evening and holidays. To relieve the queue during peak periods, the hard shoulder is temporarily opened in some expressway sections under special circumstances, allowing vehicles to drive and cooperating with speed control to improve the traffic efficiency. However, after the hard shoulder is opened, more vehicles will cross multiple lanes, intersect with the main line and ramp traffic flows, which may lead to increased lateral interference and speed difference, thus inducing new safety risks.

[0003] Existing research mainly focuses on the impact of the opening of the hard shoulder on the main line traffic flow and congestion level, and there is less quantitative evaluation of vehicle conflicts at the microscopic level. In the complex scenarios of high speed difference and multi-lane crossing, the sensitivity of traditional conflict indicators (such as TTC, PET, etc.) to sudden merging conflicts is limited, and it is difficult to accurately capture the high-risk hot spots in the weaving section. Therefore, there is an urgent need for a microscopic analysis method based on high-precision trajectory data to evaluate the conflict distribution and risk transfer before and after the "opening of the hard shoulder", and to provide a scientific basis for when and how to open the hard shoulder. Summary of the Invention

[0004] Aiming at the problems existing in the prior art, the purpose of the present invention is to provide a method for evaluating the opening risk of the hard shoulder in the weaving section of an expressway based on ITA and KDE, which can quantitatively identify and evaluate the potential conflict distribution in the weaving section before and after the opening of the hard shoulder, and provide a reference for the expressway management department to formulate hard shoulder opening strategies and supporting traffic control measures during high traffic flow periods.

[0005] To solve the above problems, the present invention adopts the following technical solutions: A method for evaluating the opening risk of the hard shoulder in the weaving section of an expressway based on ITA and KDE, comprising the following steps: Step 1: Collect and record vehicle operation data in different scenarios; Conduct aerial photography or roadside video recording at different time periods to obtain vehicle operation videos in two scenarios where the hard shoulder is not opened and opened, and obtain the speed, acceleration, ramp traffic flow, main line traffic flow, and outer lane traffic flow information of each vehicle through the vehicle operation videos; Step 2: Trajectory extraction and coordinate preprocessing; Use the Datafromsky software to identify each frame of the vehicle operation video obtained in step 1, extract the pixel-level trajectories of each vehicle, convert the pixel-level trajectories of the vehicles from pixel coordinates to the road Frenet coordinate system, and perform interpolation and filtering on the missing frames, noise, or jumps in the pixel-level trajectories to obtain trajectory sequence data with higher accuracy and more consistent resolution. At the same time, obtain intersection information based on the pixel-level trajectories of the vehicles; Step 3: Identify potential conflict risks based on ITA; After obtaining the pixel-level trajectories of the vehicles processed in step 2, calculate the ITA by analyzing the longitudinal distance and speed difference between the following vehicle and the preceding vehicle. Identify potential high-risk conflict points through ITA. When the speed difference between vehicles increases significantly and the headway or spacing is small, that is, a collision is extremely likely to occur in a short time, the ITA will rise rapidly, thus identifying potential high-risk conflict points; Step 4: Use the KDE method to spatially aggregate the conflict points; Map the conflict points with high ITA in step 3 back to the road Frenet coordinate system, use the KDE method to perform spatial aggregation analysis on the intersection information in step 2 and the distribution of high-risk conflict points in step 3, generate a two-dimensional risk heat cloud map, and obtain the KDE result; Step 5: Compare the KDE results in the scenarios where the hard shoulder is not open and open; Compare and analyze the KDE results in step 4 in two different states where the hard shoulder is not open and open, and combine the ramp flow, main line flow, and outer lane flow information obtained in step 1 to evaluate the impact degree of hard shoulder opening on the potential conflict risks in the weaving area, quantify the mitigation effect of inner lane conflicts, and the new risks of the outer lane or hard shoulder; Step 6: Determine the critical and saturated intervals of the hard shoulder in combination with the ramp flow / main line flow ratio; Under different road traffic flow levels at different times, by analyzing the scatter distribution of the ramp flow / main line flow ratio or the ramp flow / outer lane flow ratio, respectively judge whether the ramp flow / main line flow ratio or the ramp flow / outer lane flow ratio is within the corresponding threshold interval. If the ramp flow / main line flow ratio or the ramp flow / outer lane flow ratio is lower than the minimum value of the corresponding threshold interval, there is no need to open the hard shoulder; if the ramp flow / main line flow ratio or the ramp flow / outer lane flow ratio is within the corresponding threshold interval, the hard shoulder should be opened and speed limit or batch release measures should be taken as appropriate; when the ramp flow / main line flow ratio or the ramp flow / outer lane flow ratio is higher than the maximum value of the corresponding threshold interval, traffic flow control or other lane management measures need to be supplemented to prevent the further superposition of conflict risks.

[0006] Preferably, in step 1, an unmanned aerial vehicle is used for aerial photography or a roadside camera is used for roadside video recording. The shooting frame rate of the unmanned aerial vehicle or the roadside camera is not less than 30 frames per second, and the resolution is not less than 4k.

[0007] Preferably, in step 1 and step 6, different time periods include morning and evening rush hours and non-rush hours.

[0008] Preferably, in step 3, ITA is calculated by the following formula: ; Where: are respectively the longitudinal speeds of the following vehicle and the preceding vehicle in frame , with the unit of m / s; respectively represent the longitudinal positions between the following vehicle and the preceding vehicle , with the unit of m; is the length of the preceding vehicle , with the unit of m; is a correction coefficient for amplifying the sensitivity to the speed difference.

[0009] When is significantly large and is significantly small, the ITA value will increase rapidly, indicating that a collision is very likely to occur in a short time, thus indicating a high collision probability in this area.

[0010] Preferably, step 4 includes: Step 4-1: Map the conflict points with high ITA in step 3 to the road Frenet coordinate system at the same moment; Step 4-2: Perform KDE on the conflict points based on the selected bandwidth to form a two-dimensional risk heat cloud map; Step 4-3: Search for the aggregation areas exceeding the preset kurtosis threshold on the two-dimensional risk heat cloud map, define them as "high-risk hotspots", and record the corresponding coordinate ranges.

[0011] Preferably, in step 6, based on the scatter distribution analysis of the ramp flow / mainline flow ratio or the ramp flow / outer lane flow ratio at different times, the opening time of the hard shoulder is determined by respectively judging whether the ramp flow / mainline flow ratio or the ramp flow / outer lane flow ratio is within the corresponding threshold interval. The specific threshold interval of the ramp flow / mainline flow ratio is 0.12 - 0.25, and the specific threshold interval of the ramp flow / outer lane flow ratio is 0.35 - 1.0. According to the corresponding threshold interval, the flow ratio is divided into the following three intervals: The flow ratio lower than the corresponding threshold range is the low occupancy range: When the ramp flow / main line flow ratio is less than 0.12, or the ramp flow / outer lane flow ratio is lower than 0.35, the ramp flow accounts for a low proportion relative to the main line or the outer lane. Even if the hard shoulder is not opened, obvious congestion or high conflict hotspots usually will not form in the merging section. Therefore, it is not necessary to open the hard shoulder; The flow ratio equal to the corresponding threshold range is the critical increase and moderate range: When the ramp flow / main line flow ratio is between 0.12 and 0.25, or the ramp flow / outer lane flow ratio is between 0.35 and 1.0, the merging conflicts begin to increase significantly, and the hard shoulder should be opened; If this ratio has not reached the high saturation range, the hard shoulder can continue to be kept open, and speed limit or batch release and other measures can be taken as appropriate at the ramp end; The flow ratio higher than the corresponding threshold range is the high saturation range: When the ramp flow / main line flow ratio exceeds 0.25, or the ramp flow / outer lane flow ratio exceeds 1.0, the merging section has approached or reached saturation. Even if the hard shoulder is in the open state, the remaining capacity of the outer lane is difficult to effectively absorb the new flow. While opening the hard shoulder, traffic flow control or other lane management means should be supplemented to prevent the further superposition of conflict risks.

[0012] The beneficial effects of the present invention: Compared with the prior art, the advantages of the present invention are: (1) Compared with traditional conflict indicators such as TTC and PET, the ITA (Improved Accident Time Index) adopted by the present invention is more sensitive to the high speed difference situation of the vehicle speed difference, so as to accurately capture the potential conflicts in the multi-lane merging or large speed difference rear-end collision scenarios. By using drones or roadside cameras to collect the traffic flow trajectory data in the weaving area in real time, first calculate the ITA for the vehicle pairing relationship at each frame moment; then use KDE (Kernel Density Estimation) to perform spatial clustering analysis on the conflict points with high ITA in the spatio-temporal domain, so as to identify the hot spots in the weaving area. This process takes into account the severity of conflicts and the characteristics of spatial distribution, and can lay a data foundation for the subsequent impact assessment of the opening of the hard shoulder on conflict risks; (2) For the measured data (or simulation data) of the same road section during the peak period, collect the vehicle trajectories in two scenarios of "hard shoulder not opened" and "opened" respectively. By visualizing and superimposing and comparing the ITA distribution and KDE results of the two groups of data respectively, on the one hand, the conflict mitigation effect of the opening of the hard shoulder on the inner lane of the main line can be quantified; on the other hand, it can also be intuitively observed that the risk migration caused by vehicles crossing the outer lane and the hard shoulder, especially new conflict hot spots are likely to form at the ramp merging section or the connection between the outer lane and the hard shoulder. This dual effect can provide managers with a more comprehensive risk migration law; (3) To better enable the use of the hard shoulder in a dynamic traffic environment, the present invention introduces flow ratio indicators such as ramp flow / mainline flow ratio or ramp flow / outer lane flow ratio. Combining the analysis results of ITA and KDE, the flow ratio can be divided into three levels: "low ratio interval", "critical rise and moderate interval", and "high saturation interval". When the ramp flow / mainline flow ratio is between 0.12 and 0.25 (or the ramp flow / outer lane flow ratio is in the range of 0.35 to 1.0), starting the hard shoulder can significantly disperse the merging conflicts in the inner lane; if the ratio continues to rise to 0.25 or higher, it is necessary to implement ramp flow limiting or speed regulation while keeping the hard shoulder open to avoid the reappearance of high-risk agglomerations in the outer lane. This quantitative determination method is more targeted and flexible than the traditional method that relies on experience or simplified models; (4) The present invention integrates ITA (highlighting high-risk conflicts caused by vehicle speed differences) and KDE (highlighting hot spot aggregation and spatial distribution) at the micro level, and can more accurately locate high-risk merging openings or lane segments than macroscopic flow analysis in scenarios with high speed differences and multiple lane changes. At the same time, the method of the present invention has stronger adaptability to morning rush hours, extreme traffic flows, and different lane structures, and is easy to combine with the monitoring system to achieve rapid evaluation and visual presentation in actual operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a schematic diagram of the overall process of the present invention; Figure 2 is a schematic diagram of the layout of the weaving area and the hard shoulder of the ring expressway; Figure 3 is a schematic diagram of drone aerial photography and vehicle trajectory extraction; Figure 4 is a schematic diagram of vehicle trajectories in the frenet coordinate system; Figure 5 is a comparison chart of kernel density estimation results. Among them, Figure 5 (a) represents the KDE result under the condition of not opening the hard shoulder, Figure 5 (b) represents the KDE result under the condition of opening the hard shoulder; Figure 6 is a comparison chart of the spatial distribution of ITA. Among them, Figure 6 (a) represents the spatial distribution of ITA under the condition of not opening the hard shoulder, Figure 6 (b) represents the spatial distribution of ITA under the condition of opening the hard shoulder; Figure 7 is a comparison chart of the peak values of kernel density estimation at different times of the hard shoulder. DETAILED DESCRIPTION OF THE INVENTION

[0014] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0015] As Figure 1 shown, the present invention provides a technical solution: a method for evaluating the opening risk of the hard shoulder in the weaving section of expressways based on ITA and KDE, including the following steps: Step 1: Collect and record vehicle operation data under different scenarios; The present invention first selects the weaving section of the ring expressway with variable hard shoulders as the research object. As Figure 2 shown, the typical merging and diverging weaving section from Yanta to Gaoxin of the Xi'an Ring Expressway is taken as the research object. To accurately depict the road operation conditions before and after the hard shoulder is opened, DJI Air 3 drones are deployed for aerial photography during the morning and evening rush hours and non-peak hours. The drones hover stably at an altitude of 300–500 m, equipped with a Hasselblad L2D-20c 1-inch CMOS sensor camera, and output raw videos with a resolution ≥4K and a frame rate of 30 fps. During the aerial photography process, two types of time periods, namely, the hard shoulder "open" and "not open", are distinguished to ensure the accuracy and consistency of data annotation. After aerial photography, low-confidence targets are removed through multi-object detection and tracking algorithms, and information such as the speed, acceleration, ramp flow, main line flow, and outer lane flow of each vehicle is output.

[0016] Step 2: Trajectory extraction and coordinate preprocessing; As Figure 3 shown, the Datafromsky software is used to identify each frame of the vehicle operation video obtained in Step 1, and the pixel-level trajectories of each vehicle are extracted. As Figure 4 shown, the pixel coordinates of the pixel-level trajectories are then mapped to the road Frenet coordinate system (s, d), and interpolation and filtering processing are performed on the missing frames, noise, or jumps to obtain trajectory sequence data with higher accuracy and more consistent resolution. At the same time, the weaving point information is obtained according to the pixel-level trajectories of the vehicles.

[0017] Table 1: Partial vehicle data:

[0018] Step 3: Identify potential conflict risks based on ITA (Improved Accident Time Index); In the weaving area of the ring - around - the - city expressway, the confluence, divergence, and emergency lane - changing behaviors of vehicles relative to the vehicle in front or vehicles in adjacent lanes occur frequently. To quantitatively evaluate the occurrence probability of these conflicts, many researchers have introduced traffic conflict surrogate indicators to describe the approaching process of conflicts. In the field of traffic safety research, common conflict indicators such as TTC, PET, and DRAC each have their applicable scenarios and core assumptions: TTC is often used in rear - end collision scenarios where the speeds are relatively close and the speed of the following vehicle is greater than that of the preceding vehicle; PET focuses more on lateral interference or intersection scenarios; DRAC focuses on "the minimum deceleration required for the following vehicle to avoid a collision" and is sensitive to emergency braking behaviors. However, none of these indicators can encompass the dynamic evolution characteristics when "the relative speed difference is high and the vehicle changes lanes across multiple lanes". Especially after the hard shoulder is opened, the speed difference between the outer - lane vehicle speed and the main - line vehicle speed is often significantly amplified, and traditional indicators may show evaluation biases in such high - speed - difference or sudden - deceleration scenarios.

[0019] Therefore, this study selects the improved accident time index (ITA) as the main conflict surrogate indicator. To accurately capture the potential conflicts caused by the high - speed difference across multiple lanes after the hard shoulder is opened, after obtaining the vehicle pixel - level trajectories processed in step 2, ITA (improved accident time index) is calculated by analyzing the longitudinal distance and speed difference between the following vehicle and the preceding vehicle. Specifically, the present invention defines ITA using the following formula: , where: are the longitudinal speeds of the following vehicle and the preceding vehicle at frame respectively; represent the longitudinal positions of vehicles respectively; is the length of the preceding vehicle ( vehicle); is a correction coefficient for amplifying the sensitivity to the speed difference. When is significantly large and is significantly small, ITA will increase rapidly, indicating that a collision is very likely to occur in a short time, thus indicating a relatively high collision probability in this area and identifying potential high - risk conflict points. As Figure 6 shown, Figure 6 (a) represents the ITA spatial distribution under the condition of the unopened hard shoulder, Figure 6 (b) represents the ITA spatial distribution under the condition of the opened hard shoulder.

[0020] Step 4: Use the KDE (kernel density estimation) method to perform spatial aggregation on conflict points: After extracting the coordinates of the weaving points based on the vehicle trajectories, KDE (Kernel Density Estimation) can directly construct a relatively smooth density function according to the distribution of sample points in space or spatio-temporal space. Different from the traditional parametric distribution fitting, KDE does not require a preset distribution type, nor does it rely on the specific parameter estimation process, so it is applicable to many hotspot detections or cluster identifications based on real observation data in the field of road traffic. The basic starting point of this method is: assuming a certain point in space or where the density value is jointly determined by the number and distance of the surrounding sample points. Specifically, for one-dimensional kernel density estimation, if there are independent and identically distributed samples from an unknown density function , then the estimated value at a certain point can be expressed as:

[0021] where is called the kernel function, is the bandwidth, is the estimated point and the distance between the sample point , and n is the total number of event points. The magnitude of the kernel function value reflects the concept that "the closer to the center point, the greater the weight; the farther away, the smaller the contribution to the density of this point". For the traffic data of the ring expressway, when we have the distribution of vehicle positions or conflict events on a one-dimensional line, one-dimensional kernel density estimation can obtain the aggregation distribution of events on the longitudinal scale of the road, and the maximum value of its clustering points is the peak value we need to find.

[0022] In practical applications, since road networks or weaving areas often involve horizontal and vertical (even temporal) distribution situations, two-dimensional or even multi-dimensional kernel density estimation is required for analysis. For example, in the two-dimensional scenario of the weaving area, let the bandwidth vector be , and assume the sample data is , then the kernel density estimation of any estimated point can be written as:

[0023] where is the kernel function, is the total number of samples; represents the horizontal and vertical distances between this estimated point and each sample point.

[0024] In the microscopic traffic flow analysis of the weaving area of the ring expressway, the behaviors of vehicle merging, diverging, and lane-changing will form a relatively concentrated interference zone in space and time. If this interference zone is superimposed with a large speed difference or a small longitudinal spacing, it is more likely to evolve into a potential conflict or an actual collision risk. To identify these high-risk areas, kernel density estimation is introduced to conduct a spatial aggregation analysis of the position distribution of weaving behaviors. As Figure 5 shown Figure 5 (a) represents the KDE result under the condition of the closed hard shoulder, Figure 5 (b) represents the KDE result under the condition of the open hard shoulder.

[0025] Step 5: Compare the KDE results under the scenarios of the closed and open hard shoulders; Compare and analyze the KDE results in step 4 under different hard shoulder states, and combine the ramp flow, main line flow, and outer lane flow information obtained in step 1 to evaluate the influence degree of the open hard shoulder on the potential conflict risk in the weaving area, and quantify the mitigation effect of the inner lane conflict and the new risks of the outer lane or the hard shoulder.

[0026] From Figure 5 and Figure 6It can be seen that when the hard shoulder is not open, the weaving points are mainly concentrated in the first and second lanes (inner lanes) of the main line, especially forming several obvious high-density "hot spots" near the merging and diverging points. This is because the lateral space of the main line is limited during this period, and vehicles tend to frequently change lanes in the main line lanes when entering or leaving the weaving area, resulting in the mutual interference between vehicles being concentrated in this area. In addition, although there is also a certain weaving phenomenon in the third lane (outer lane) of the main line, compared with the inner lane, its density value is significantly lower, indicating that the merging behavior of vehicles in the outer lane is relatively less and more dispersed. After the hard shoulder is opened, there is a certain "outward shift" and "dispersion" phenomenon in the spatial distribution of the weaving hot spots: on the one hand, the high-density area in the first and second lanes of the main line weakens, and there is a trend of density reduction or hot spot range reduction locally; on the other hand, a new high-density weaving area appears in the third lane, the collector-distributor lane and the hard shoulder area. Especially during the morning peak congestion period, some vehicles choose to drive on the hard shoulder or the outer lane due to the saturation of the main line traffic flow, resulting in a significant increase in the merging and diverging behaviors of these lanes. It should be noted that in the outermost hard shoulder area, due to the different vehicle entry and exit paths and the large speed difference, the local kernel density value is higher than that in the non-open scenario. While the opening of the hard shoulder alleviates the congestion and weaving phenomenon in the inner lane, it also transfers some of the merging and diverging behaviors to the outer lane and even the hard shoulder, having a new impact on the spatial risk distribution of the overall weaving area. There is a strong correlation between the distribution of ITA and the kernel density estimation. When the hard shoulder is not open, due to the relatively large number of main line lanes, the distribution of high risk values is more concentrated in the first and second lanes. At the same time, the risk area of the collector-distributor lane is not at the merging and diverging points, but in the area where the main line intersects with the merging and diverging vehicles, indicating that there is also a certain conflict risk between the diverging vehicles and the merging vehicles on the ramp. After the hard shoulder is opened, although the risk in some areas is alleviated, at the junction of the third lane and the hard shoulder, the ITA increases significantly, which corresponds to the "outward shift" hot spot of the kernel density estimation, indicating that the new merging and diverging conflicts are mainly concentrated in the outer lane and the hard shoulder area. In addition, there is a phenomenon that the ITA is not completely consistent with the weaving point density in some sections. Especially when in the first lane or closer to the inner lane, the ITA does not increase significantly with the number of weaving points. The possible reason is that these vehicles often complete the deceleration or lane-changing actions after merging into the main line, the conflict process is short and the risk is reduced, having little impact on the final ITA.

[0027] Combining the kernel density estimation in Step 3 and the risk situation distribution in Step 4, it can be seen that the opening of the hard shoulder has a certain effect on alleviating the congestion and weaving risks in the inner lane, but at the same time, new high-risk areas are formed in the outer lane and the hard shoulder area. The main reasons are as follows: When the hard shoulder is opened, some vehicles choose to transfer to the hard shoulder or the outermost lane, thus reducing the traffic flow and density of the main line lanes. However, due to the original design intention of the hard shoulder not being a conventional driving lane, the lateral width and safety facility configuration are limited, resulting in a large change in the speed difference and merging time interval of vehicles in this area, and the local conflict risk increases. During the morning peak period, if the vehicle speed fluctuates violently, such multi-lane crossings are more likely to cause unsafe events. At the same time, since most main line vehicles use the hard shoulder and the collector-distributor lane to accelerate and try to merge back into the main line, the congestion value and risk value increase abnormally in the merging area. The kernel density estimation shows that the merging and diverging behaviors are significantly concentrated near the hard shoulder or the collector-distributor lane. If the traffic flow exceeds the safety carrying threshold of the hard shoulder, the lateral interference and longitudinal speed difference between vehicles will accumulate rapidly, resulting in an increase in ITA, thereby evaluating the impact degree of the hard shoulder opening on the potential conflict risks in the weaving area.

[0028] Step 6: Determine the critical and saturated intervals of the hard shoulder in combination with the ramp / main line flow ratio: As Figure 7 shown, based on the multi-period measured data analysis, the present invention visually compares the peaks of the kernel density estimation hotspots in the two scenarios of "hard shoulder not opened" and "hard shoulder opened", and it can be observed that: When the hard shoulder is opened, the conflicts in the inner lane usually ease; However, once the traffic flow in the outer lane or the hard shoulder increases significantly, and the ramp flow / main line flow ratio or the ramp / outer lane ratio exceeds the corresponding threshold (such as 0.25 or 1.0), new conflict aggregation points are likely to form in the outer lane or the hard shoulder. Therefore, the present invention divides these ratio intervals into "low occupancy interval", "critical rising and moderate interval" and "high saturation interval", and it is recommended to actively open the hard shoulder to divert traffic in the critical interval. Once entering the high saturation interval, ramp flow limiting or speed control measures must be supplemented, otherwise secondary conflicts will occur on the outer side. If the traffic flow continues to surge or there are breakdown vehicles on the hard shoulder, the hard shoulder needs to be temporarily closed to retain the emergency function and reduce additional risks.

[0029] The application effect of the present invention is as follows: Through the actual measurement in the weaving area of the ring expressway interchange, it is found that the method of the present invention has a high sensitivity to the high speed difference and multi-lane crossing interference, which is superior to traditional conflict indicators such as TTC and PET. When the ramp flow / main line flow ratio is about 0.15 - 0.20, the opening of the hard shoulder significantly disperses the inner merging conflicts; when the ratio is greater than 0.25, ramp flow limiting needs to be coordinated to prevent the outer lane or the hard shoulder from being congested again and causing high risks.

[0030] Determination of the opening time of the hard shoulder: When further analyzing the scatter distribution of the ramp flow / mainline flow ratio or the ramp flow / outer lane flow ratio, it can be divided into three scenarios: "low occupancy interval", "critical rise and moderate interval", and "high saturation interval": When the ramp flow ratio is relatively low (e.g., less than 0.12), it is usually not necessary to open the hard shoulder and serious congestion will not occur; When the ratio is between 0.12 and 0.25, the hard shoulder should be opened in a timely manner, and lane speed limits or batch releases should be adopted as appropriate; When the ratio continues to increase (greater than 0.25 or greater than 1.0), it indicates that the confluence section has approached saturation, and more strict traffic flow control or other lane management means should be supplemented to prevent the further superposition of conflict risks.

[0031] If a vehicle breakdown occurs when the hard shoulder is opened, it must be immediately converted into an emergency lane to ensure the rescue passage. The present invention can also be linked with microscopic simulation for real-time or off-line evaluation of larger-scale road sections, and has high popularization and application value.

[0032] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A risk assessment method for hard shoulder opening in highway weaving areas based on ITA and KDE is characterized by: The following steps are involved: Step 1: Collect and record vehicle operation data in different scenarios; Aerial photography or roadside video recording is performed at different time periods to obtain vehicle operation videos in two scenarios: hard shoulder is closed and open. The speed, acceleration, ramp flow, main line flow, and outer lane flow information of each vehicle are obtained through the vehicle operation videos. Step 2: trajectory extraction and coordinate preprocessing; Use Datafromsky software to identify the vehicle running video obtained in step 1 frame by frame, extract the pixel-level trajectory of each vehicle, convert the pixel-level trajectory of the vehicle from the pixel coordinate to the road Frenet coordinate system, and perform interpolation and filtering on the pixel-level trajectory, and obtain the intersection point information according to the pixel-level trajectory of the vehicle; Step 3: Identify potential conflict risks based on ITA; After obtaining the pixel-level trajectory of the vehicle processed in step 2, ITA is calculated by analyzing the longitudinal distance and speed difference between the rear vehicle and the front vehicle, and potential high-risk conflict points are identified through ITA; Step 4: Use KDE method to spatially cluster conflict points; Map the conflict points with high ITA in step 3 back to the road Frenet coordinate system, use the KDE method to perform spatial clustering analysis on the interweaving point information in step 2 and the distribution of high-risk conflict points in step 3, generate a two-dimensional risk heat cloud map, and obtain the KDE result; Step 5: Compare the KDE results of the hard shoulder in the closed and open scenarios; The KDE results in step 4 are compared and analyzed in two different states: the hard shoulder is not open and the hard shoulder is open. Combined with the ramp flow, main line flow and outer lane flow information obtained in step 1, the impact of the hard shoulder opening on the potential conflict risk in the weaving area is evaluated, and the mitigation effect of the inner lane conflict and the additional risk of the outer lane or hard shoulder are quantified; Step 6: Determine the critical and saturated intervals of the hard shoulder based on the ramp flow / mainline flow ratio; Under the road flow level in different time periods, by analyzing the scatter distribution of the ramp flow / mainline flow ratio or the ramp flow / outer lane flow ratio, it is judged whether the ramp flow / mainline flow ratio or the ramp flow / outer lane flow ratio is within the corresponding threshold interval. If the ramp flow / mainline flow ratio or the ramp flow / outer lane flow ratio is lower than the minimum value of the corresponding threshold interval, there is no need to open the hard shoulder; if the ramp flow / mainline flow ratio or the ramp flow / outer lane flow ratio is within the corresponding threshold interval, the hard shoulder should be opened; when the ramp flow / mainline flow ratio or the ramp flow / outer lane flow ratio is higher than the maximum value of the corresponding threshold interval, flow control is required.

2. The risk assessment method for hard shoulder opening in weaving areas of highways based on ITA and KDE according to claim 1 is characterized in that: In step 1, a drone is used for aerial photography or a roadside camera is used for roadside photography. The shooting frame rate of the drone or roadside camera is not less than 30 frames per second, and the resolution is not less than 4k.

3. The risk assessment method for hard shoulder opening in weaving areas of highways based on ITA and KDE according to claim 1 is characterized in that: In step 1 and step 6, different time periods include morning and evening peak hours and non-peak hours.

4. The risk assessment method for hard shoulder opening in weaving areas of highways based on ITA and KDE according to claim 1 is characterized in that: In step 3, ITA is calculated using the following formula: ； in: Rear car With the front car In frame Longitudinal velocity, in m / s; Respectively indicate the following vehicles With the front car The longitudinal position between the two, in m; For the front car The length of the unit is m; It is the correction factor for amplifying the sensitivity of speed difference.

5. The risk assessment method for hard shoulder opening in weaving areas of highways based on ITA and KDE according to claim 1 is characterized in that: Step 4 includes: Step 4-1: Map the conflict points with high ITA in step 3 to the road Frenet coordinate system at the same time; Step 4-2: Perform KDE on the conflict points based on the selected bandwidth to form a two-dimensional risk heat cloud map; Step 4-3: Find the clustered areas that exceed the preset kurtosis threshold on the two-dimensional risk heat cloud map, define them as "high-risk hotspots", and record the corresponding coordinate range.

6. The risk assessment method for hard shoulder opening in weaving areas of highways based on ITA and KDE according to claim 1 is characterized in that: In step 6, based on the scatter point distribution analysis of the ramp flow / mainline flow ratio or the ramp flow / outer lane flow ratio in different time periods, the opening timing of the hard shoulder is determined by judging whether the ramp flow / mainline flow ratio or the ramp flow / outer lane flow ratio is within the corresponding threshold interval. The specific threshold interval of the ramp flow / mainline flow ratio is 0.12-0.25, and the specific threshold interval of the ramp flow / outer lane flow ratio is 0.35-1.

0. The flow ratio is divided into the following three intervals according to the corresponding threshold intervals: The interval below the corresponding threshold is a low proportion interval: when the ratio of ramp flow to main line flow is less than 0.12, or the ratio of ramp flow to outer lane flow is less than 0.35, the proportion of ramp flow relative to the main line or outer lane is low, and there is no need to open the hard shoulder; The corresponding threshold interval is the critical rise and moderate interval: when the ratio of ramp flow / mainline flow is between 0.12 and 0.25, or the ratio of ramp flow / outer lane flow is between 0.35 and 1.0, the hard shoulder should be opened; The interval above the corresponding threshold is a high saturation interval: when the ramp flow / main line flow ratio exceeds 0.25, or the ramp flow / outer lane flow ratio exceeds 1.0, flow control should be provided while opening the hard shoulder.

Images